差分电路终端接法比较

DS26C31T,DS26C32AM,DS26C32AT, DS26F32MQML,DS26LS31C,DS26LS31M, DS26LS32AC,DS26LS32C,DS26LS33M, DS26LV31QML,DS26LV31T,DS26LV32AQML, DS26LV32AT,DS34C86T,DS34C87T,DS34LV86T, DS34LV87T Application Note 903 A Comparison of Differential Termination Techniques Literature Number: SNLA034A A Comparison of Differential Termination Techniques Introduction Transmission line termination should be an important con- sideration to the designer who must transmit electrical sig- nals from any point A to any point B. Proper line termination becomes increasingly important as designs migrate towards higher data transfer rates over longer lengths of transmis- sion media. However, the subject of transmission line termi- nation can be somewhat confusing since there are so many ways in which a signal can be terminated. Therefore, the advantages and disadvantages of each termination option are not always obvious. The purpose of this application note is to remove some of the confusion which may surround signal termination. This dis- cussion, however, will focus attention upon signal termina- tion only as it applies to differential data transmission over twisted pair cable. Common differential signal termination techniques will be presented and the advantages and disad- vantages of each will be discussed. Each discussion will also include a sample waveform gener- ated by a setup consisting of a function generator whose signals are transmitted across a twisted pair cable by a differential line driver and sensed at the far end by a differ- ential line receiver. This application note will specifically address the following differential termination options: • Unterminated • Series/Backmatch • Parallel • AC • Power (Failsafe) • Alternate Failsafe • Bi-Directional For the purposes of discussion, popular TIA/EIA-422 drivers and receivers, such as the DS26LS31 and DS26LS32A, will be used to further clarify differential termination. Unterminated The selection of one termination option over another is of- tentimes dictated by the performance requirements of the application. The selection criteria may also hinge upon other factors such as cost. From this cost perspective the option of not terminating the signal is clearly the most cost effective solution. Consider Figure 1, where a DS26LS31 differential driver and a DS26LS32A differential receiver have been connected (using a twisted pair cable) together without a termination element. Because there is no signal termination element, the DS26LS31 driver’s worst case load is the DS26LS32A receiver’s minimum input resistance. Since, TIA/EIA-422-A (RS-422) standard defines the DS26LS32A’s minimum input resistance to be 4 kΩ, the driver’s worst case load, as seen in Figure 1, is then 4 kΩ. In the unterminated configuration, the DS26LS31 driver is only required to source a minimal amount of current in order to drive a signal to the receiver. This minimal DC current sourcing requirement in turn minimizes the driver’s on chip power dissipation. In addition, the 4 kΩ driver output load results in a higher driver output swing (than if the driver was loaded with 100Ω) which in turn increases DC noise margin. This increase in noise margin further diminishes the possi- bility that system noise will improperly switch the receiver. To be sure that there is no confusion, noise margin is defined as the difference between the minimum driver output swing and the maximum receiver sensitivity. On the other hand, if a receiver was used which complies to TIA/EIA-485 (RS-485), the resulting noise margin would be even greater. This is because the minimum input resistance of an RS-485 re- ceiver must be greater than 12 kΩ as compared to 4 kΩ for an RS-422 receiver. The absence of a termination element at the DS26LS32A’s inputs also guarantees that the receiver output is in a known logic state when the transmission line is in the idle or open line state (receiver dependent). This condition is commonly referred to as open input receiver failsafe. This receiver failsafe (Note 1) bias is guaranteed by internal pull up and pull down resistors on the positive and negative receiver inputs, respectively. These pull up and pull down resistors bias the input differential voltage (VID) to a value greater than 200 mV when the line is, for example, idle (un-driven). This bias is significant in that it represents the minimum guaran- teed VID required to switch the receiver output into a logic high state. Note: A complete discussion of receiver failsafe can be found in Application Note 847 (AN-847). There are, however, some disadvantages with an untermi- nated cable. The most significant effect of unterminated data transmission is the introduction of signal reflections onto the transmission line. Basic transmission line theory states that a signal propagating down a transmission line will be re- flected back towards the source if the outbound signal en- counters a mismatch in line impedance at the far end. In the case of Figure 1, the mismatch occurs between the charac- teristic impedance of the twisted pair (typically 100Ω) and the 4 kΩ input resistance of the DS26LS32A. The result is a signal reflection back towards the driver. This reflection then encounters another impedance mismatch at the driver out- puts which in turn generates additional reflections back to- ward the receiver, and so on. The net result is a number of reflections propagating back and forth between the driver and receiver. These reflections can be observed in Figure 2. 01189801 FIGURE 1. Unterminated Configuration National Semiconductor Application Note 903 Joe Vo August 1993 A Com parison ofDifferentialTerm ination Techniques AN-903 © 2002 National Semiconductor Corporation AN011898 www.national.com Unterminated (Continued) The main limitation of unterminated signals can be clearly seen in Figure 2. A positive reflection is generated when the signal encounters the large input resistance of the receiver. These reflections propagate back and forth until a steady state condition is reached after several round trip cable delays. The delay is a function of the cable length and the cable velocity. Figure 2 shows that the reflections settle after three round trips. To limit the effect of these reflections, unterminated signals should only be used in applications with low data rates and short driving distances. The data being transmitted should, therefore, not make any transitions until after this steady state condition has been reached. A low data rate ensures that reflections have suffi- cient time to settle before the next signal transition. At the same time, a short cable length ensures that the time re- quired for the reflections to settle is kept to a minimum. The low data rate and short cable length dictated by the lack of termination is probably the most significant shortcoming of the unterminated option. Low speed is generally characterized to be either signalling rates below 200 kbits/sec or when the cable delay (the time required for an electrical signal to transverse the cable) is substantially shorter than the bit width (unit interval) or when the signal rise time is more than four times the one way propagation delay of the cable (i.e., not a transmission line). As a general rule, if the signal rise time is greater than four times the propagation delay of the cable, the cable is no longer considered a transmission line. It should be mentioned that most differential data transmis- sion applications provide for some kind of signal termination. This is because most differential applications transmit data at relatively high transfer rates over relatively long distances. In these type of applications, signal termination is critically important. If the application only requires low speed opera- tion over short distances, an unterminated transmission line may be the simplest solution. Series Termination Another termination option is popularly known as either se- ries or backmatch termination. Figure 3 illustrates this type of termination. The termination resistors, RS, are chosen such that their value plus the impedance of the driver’s output equal the characteristic impedance of the cable. Now as the 01189802 FIGURE 2. Unterminated Waveforms AN -9 03 www.national.com 2 Series Termination (Continued) driven signal propagates down the transmission line an im- pedance mismatch is still encountered at the far end of the cable (receiver inputs). However, when that signal propagates back to the driver the reflection is terminated at the driver output. There is only one reflection before the driven signal reaches a steady state condition. How long it takes for the driven signal to reach steady state is still dependent upon the length of cable the signal must traverse. As with the unterminated option the driver power dissipation is still minimized due to the light loading presented by the 4 kΩ receiver input resistance. The driver loading remains unchanged from the unterminated option. In both cases the driver is effectively loaded with the receiver’s input impedance. DC noise margin has again increased and the open input receiver failsafe feature is still supported for idle and open line conditions. There are three major disadvantages in using series termi- nation. First, the driver output impedances can vary, due to normal process variations, from one manufacturer to another and from one driver load to another. Should there be a problem which involves replacing line drivers, there is a chance that the designer might have to rework the board in order to ensure that the RS matches the new driver’s output impedance. Second, series termination is commonly limited to only point to point applications. Consider the following example. If a second receiver (multidrop application) was located halfway between the driver and receiver at the far end of the cable, the noise margin seen by the middle receiver would change between the incident signal and the reflected signal. Such a problem would not exist in a point to point application where only one receiver is used with one driver. Third, there is still an impedance mismatch at the receiver inputs. Again, this mismatch is caused by a signal propagat- ing down a 100Ω cable suddenly encountering a 4 kΩ re- ceiver input resistance. This impedance mismatch will con- tinue to cause reflections on the transmission line as illustrated in Figure 4. Notice the reflections which result when the driven signal encounters an impedance mismatch at the receiver input. The reflection propagates back to the driver and is some- what terminated by the driver’s output impedance. The re- flected signal is terminated because the combined imped- ance of the series resistor (RS) and the driver’s output impedance comes close to matching the characteristic im- pedance of the cable. In contrast with Figure 2’s untermi- nated signal waveform, the waveform seen in Figure 4 is characterized by only one reflection. In all it will take the signal one round trip cable delay to be reflected back towards the signal source. Since all reflec- tions should be allowed to settle before the next data tran- 01189803 FIGURE 3. Series Termination Configuration 01189804 FIGURE 4. Series Termination Waveforms AN-903 www.national.com3 Series Termination (Continued) sition (to maintain data integrity), it is imperative that the round trip cable delay be kept much less than the time unit interval (TUI — defined to be the minimum bit width or the “distance” between signal transitions). In other words, series termination should be limited to applications where the cable lengths are short (to minimize round trip cable delays) and the data rate is low (to maximize the TUI). And to a lesser degree, the series termination option may not be the ideal choice from a cost perspective in that it requires two addi- tional external components. Parallel Termination Parallel termination is arguably one of the most prevalent termination schemes today. In contrast to the series termi- nation option, parallel termination employs a resistor across the differential lines at the far (receiver) end of the transmis- sion line to eliminate all reflections. See Figure 5. Eliminating all reflections requires that RT be selected to match the characteristic impedance (ZO) of the transmission line. As a general rule, however, it is usually better to select RTsuch that it is slightly greater than ZO. Over-termination tends to be more desirable than under-termination since over-termination has been observed to improve signal qual- ity. RT is typically chosen to be equal to ZO. When over-termination is used RT is typically chosen to be up to 10% larger than ZO. The elimination of reflections permits higher data rates over longer cable lengths. Keep in mind, however, that there is an inverse relationship between data rate and cable length. That is, the higher the data rate the shorter the cable and conversely the lower the data rate the longer the cable. Higher data rates and longer cable lengths translate simply into smaller TUI’s and longer cable delays. Unlike series termination where high data rates and long cable lengths can negatively impact data integrity, parallel termination can effectively remove all reflections; thereby removing all concerns about reflections interfering with data transitions. See Figure 6. 01189805 FIGURE 5. Parallel Termination Configuration AN -9 03 www.national.com 4 Parallel Termination (Continued) As seen in Figure 6 both driver output and receiver input signals are free of reflections. Such results make parallel termination optimal for use in either high speed (10 Mb/s), or long cable length (up to 4000 feet), applications. Another benefit the parallel termination provides is that both point to point and multidrop applications are supported. Re- call that multidrop is defined as a distribution system com- posed of one driver and up to ten receivers spread out along the cable as defined in the TIA/EIA-422 standard. The par- allel termination is located at the far end (opposite the driver) of the cable and effectively terminates the signal at that location, preventing reflections. There are also disadvantages to parallel termination. Let’s examine these disadvantages as they pertain to multidrop configurations. An intrinsic assumption to multidrop opera- tion is that stub lengths, as measured by “I” in Figure 5, are minimized. Despite the fact that all receivers are effectively terminated with RT, long stub lengths will once again reintro- duce impedance mis-matches and reflections. So while par- allel termination may remove reflections and permit multi- drop configurations, it does place a restriction upon the stub lengths associated with these other receivers. Typically stubs should be kept to less that 1⁄4 of the drivers rise time in length to minimize transmission line effects, and reflections. TIA/EIA-422-A standard does recommend a 100Ω resistor to be used when the differential line is parallel terminated. Therefore, applications which use a TIA/EIA-422-A driver such as the DS26LS31 or DS26C31 are commonly termi- nated with 100Ω at the far end of the twisted pair cable. While the 100Ω parallel termination eliminates all reflections, the power dissipated by the driver will increase substantially with the addition of this resistor. This increased driver power dissipation is a major disadvantage of parallel termination. The absence of this termination resistor keeps driver power dissipation low for unterminated and series terminated driv- ers and is a major advantage of these two termination op- tions. Noise margin will also decrease with parallel termination. The relatively light loading (4 kΩ) of unterminated and series terminated drivers led to larger driver output swings. The heavier driver Ioad (typically 100Ω) brought on by parallel termination reduces the driver’s output signal swing. How- ever, even with this reduction, there is ample noise margin left to ensure that the receiver does not improperly switch. Recall the discussion earlier about receiver failsafe with the unterminated and series options. In both cases, open input receiver failsafe operation was guaranteed because of inter- nal circuitry (receiver dependent) which biases the differen- 01189806 FIGURE 6. Parallel Termination Waveforms AN-903 www.national.com5 Parallel Termination (Continued) tial input voltage (VID) to a value greater than its differential threshold. Since the resulting bias voltage at the receivers inputs (VID), is greater than +200 mV, the output of the DS26LS32A receiver remains in a stable HIGH state. Unlike unterminated and series options, parallel termination cannot support open input receiver failsafe when the transmission line is in the idle state. This shortcoming of parallel termina- tion is discussed in much greater detail later in the section which describes power and alternate failsafe termination (see AN-847 for more of information on failsafe biasing differential buses). AC Termination The effectiveness of parallel termination is oftentimes coun- tered by increased driver power dissipation and receiver failsafe concerns. The DC loop current required by the ter- mination resistor, RT (see Figure 5), is often too large in order to be useful for power conscious applications or for seldomly switched control lines. In asynchronous applica- tions, parallel termination’s is not able to guarantee receiver failsafe during idle bus states which in turn makes the sys- tem susceptible to errors such as false start bits and framing errors. The primary reason for the AC termination, however, grew out of the need for effective transmission line termina- tion with minimal DC loop current. A representation of an AC terminated differential line is shown in Figure 7. The value of RT generally ranges from 100Ω–150Ω (cable ZO dependent) and is selected to match the characteristic impedance (ZO) of the cable. CT, on the other hand, is selected to be equal to the round trip delay of the cable divided by the cable’s ZO. EQ1: CT ≤ (Cable round trip delay) / ZO For this example: Cable Length = 100 feet Velocity = 1.7 ns/foot Char. Impedance = 100Ω Therefore, CT ≤ (100 ft x 2 x 1.7 ns/ft)/100Ω or ≤ 3,400 pF. Further, the resulting RC time constant should be less than or equal to 10% of the unit interval (TUI). In the example provided the maximum switching rate therefore should be less than 300 kHz. This termination should now behave like a parallel termination during transitions, but yield the ex- panded noise margins during steady state conditions. See Figure 8. Figure 8 illustrates the tradeoff between parallel terminated and unterminated signals. There are no major reflections and driver power dissipation is reduced at the expense of a low pass filtering effect which essentially limits the applica- tion of AC termination to low speed control lines. Note that the frequency of the driven signal in Figure 8 is 300 kHz whereas it was 500 kHz for the other plots. This was done to maintain the ratio between bit time and the RC time constant. The draft revision of RS-422-A will include AC termination as an alternative to paralleI termination. 01189807 FIGURE 7. AC Termination Configuration AN -9 03 www.national.com 6 AC Termination (Continued) The waveforms in Figure 8 should be viewed together with the following brief explanation of how AC termination works. When the driven signal transitions from one logic state to another, the capacitor CT behaves as a short circuit and consequently, the load presented to the driver is essentially RT. However, once the driven signal reaches its intended levels, either a logic HIGH or logic LOW, CT will behave as an open circuit. DC loop current is now blocked. The driver power dissipation will then decrease. The load presented to the driver also decreases. This is due to the fact that the driver is n